Closed Bug 927255 Opened 12 years ago Closed 5 years ago

Some heap-unclassified due to IPC strings(?)

Categories

(Core :: IPC, defect, P4)

ARM
Gonk (Firefox OS)
defect

Tracking

()

RESOLVED WONTFIX

People

(Reporter: n.nethercote, Unassigned)

Details

(Keywords: perf, Whiteboard: [c=memory p=5 s= u=] [MemShrink:P2])

Attachments

(2 files)

+++ This bug was initially created as a clone of Bug #919864 +++ Attachment 811297 [details] from bug 919864 shows 43 MiB of unreported strings, related somehow to IPC code. I don't know where the memory allocated at that point ends up being stored, which is the information necessary to write a reporter for it.
My understanding was that they were being used as the backing store for blobs, or something. So maybe one of the categories of blob things that aren't yet accounted for.
Whiteboard: [MemShrink] → [c=memory p= s= u=] [MemShrink]
Whiteboard: [c=memory p= s= u=] [MemShrink] → [c=memory p= s= u=] [MemShrink:P2]
Priority: P1 → P2
Assignee: nobody → wcosta
Status: NEW → ASSIGNED
Whiteboard: [c=memory p= s= u=] [MemShrink:P2] → [c=memory p=5 s= u=] [MemShrink:P2]
Hi Wander! The starting point for this bug is to see if it still happens. It seems like it's intermittent. Instructions for DMD are at https://wiki.mozilla.org/Performance/MemShrink/DMD. Please ask if you have any questions.
The "interesting" log starts at line 345.
(In reply to Comment 2) Hi Nicholas! It took more time than I was expecting but I could confirm the bug is still there. Attachment 8460557 [details] has full compressed logs, it is almost 80000 lines long, due to a lot of graphics related code reports. I am not sure if this is expected. The IPC string report starts at line 345. For convenience, Attachment 8460563 [details] contains an uncompressed log with only the IPC string report.
Full DMD logs tend to be pretty long, so that's expected :) The IPC stuff is still there, but the size is much smaller than the one reported in comment 0 -- only 114 KiB instead of 43 MiB. It would still be nice to have a memory reporter for this memory so that it shows up in the memory reports, just in case it does occasionally get much larger. The tricky part with doing that is working out *where* this particular memory is stored -- I tried and failed when I looked at this a while ago. https://wiki.mozilla.org/Memory_Reporting has some documentation on how to write a memory reporter. It would be also worthwhile to look at xpcom/base/nsIMemoryReporter.idl.
Downgrading to P4 based on njn's comment 6: (In reply to Nicholas Nethercote [:njn] from comment #6) > > The IPC stuff is still there, but the size is much smaller than the one > reported in comment 0 -- only 114 KiB instead of 43 MiB... >
Severity: critical → normal
Priority: P2 → P4
Summary: Lots of heap-unclassified due to IPC strings(?) → Some heap-unclassified due to IPC strings(?)
In both cases the memory is coming from DOM blobs.
I thought the problem was when you take a screenshot, but it actually fires when you run the gallery app.
Assignee: wcosta → nobody
Moving to IPC, not sure if this is still necessary. It's possible we added reporters in the past 3 years.
Component: Gaia::System → IPC
Product: Firefox OS → Core
No assignee, updating the status.
Status: ASSIGNED → NEW
No assignee, updating the status.

No action in several years, closing.

Status: NEW → RESOLVED
Closed: 5 years ago
Resolution: --- → WONTFIX
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Created:
Updated:
Size: